a theory of tonal hierarchies in music - the music cognition

51

3.1 Introduction

One of the most pervasive structural principles found in music historically and cross-culturally is a hierarchy of tones. Certain tones serve as reference pitches; they are stable, repeated frequently, are emphasized rhythmically, and appear at structurally important positions in musical phrases. The details of the hierarchies differ across styles and cultures. Variation occurs in the particular intervals formed by pitches in the musical scale and the hierarchical levels assigned to pitches within the scale. This variability suggests that an explanation for how these hierarchies are formed cannot be derived from invariant acoustic facts, such as the harmonic structure (overtones) of complex tones. Rather, the evidence increasingly suggests that these hierarchies are products of cognition and, moreover, that they rely on fundamental psychological principles shared by other domains of perception and cognition.

In this chapter, a theory of tonal hierarchies is presented that rests upon three interrelated propositions. The first is that tonal hierarchies have psychological reality. The first is that tonal hierarchies have psychological reality – that is, they are rep-resented cognitively and play that is, they are represented cognitively and play a central role in how musical sequences are perceived, organized, and remembered and in how expectations are formed during listening. This proposition implies that effects of tonal hierarchies should surface in a variety of empirical measures, such as direct judgments of musical structure, memory errors, and neurophysiological measures. The second proposition is that the tonal hierarchies are also musical facts. As such, it is expected that these hierarchies will manifest in the way music is written and how its structure is codified in music theory. Tonal hierarchies should be evident in the musical surface and characterize otherwise diverse musical styles.

C.L. Krumhansl (*) Department of Psychology, Cornell University, Ithaca, NY 14853, USA e-mail: [email protected]

L.L. Cuddy (*) Department of Psychology, Queen’s University, Kingston, Ontario K7L 3N6, Canada e-mail: [email protected]

Chapter 3A Theory of Tonal Hierarchies in Music

Carol L. Krumhansl and Lola L. Cuddy

M.R. Jones et al. (eds.), Music Perception, Springer Handbook of Auditory Research 36,DOI 10.1007/978-1-4419-6114-3_3, © Springer Science+Business Media, LLC 2010

52 C.L. Krumhansl and L.L. Cuddy

The third proposition is that statistically frequent patterns in the music should, in most cases, be reliable guides to the listener for abstracting the tonal hierarchy. This proposition would predict that listeners are able to orient relatively rapidly to the style-appropriate tonal hierarchy, and that perceptual judgments should converge with statistical distributions of tones and tone combinations. The psychological challenge with which the present chapter is concerned is the isolation, direct mea-surement, and quantification of tonal hierarchies.

3.2 Tonal Hierarchy

Tonal hierarchy refers to both a fundamental theoretical concept in describing musical structure and a well-studied empirical phenomenon. As a theoretical concept, the essential idea is that a musical context establishes a hierarchy of tones. Certain musical tones are more prominent, stable, and structurally significant than others, thus yielding a hierarchical ordering of tones. For Western tonal-harmonic music (the prevalent music of the eighteenth and nineteenth centuries) the first tone of the scale (the tonic) is said to head the hierarchy. This tone is followed by the fifth and third scale degrees (the dominant and the mediant, respectively), the other (major or minor) scale tones, and finally the nonscale tones. This hierarchy reflects the influence of triadic (chord) structure in this style, in which consonant chords predominate (for summaries of these elementary aspects of Western music theory, see Handel 1989; Krumhansl 1990a; Patel 2008; Thompson 2008).

As an empirical phenomenon, tonal hierarchy has been extensively investigated in psychological experiments. These investigations are motivated by two general objectives. The first is to test the psychological reality of music-theoretical descrip-tions. Do the observations made by music theorists about tonal hierarchies have consequences for how musical pitch is perceived and remembered? The second objective is to locate the observed empirical phenomenon within the broader theo-retical and methodological framework of psychology. Through what processes are tonal hierarchies internalized? What relationship do tonal hierarchies have to other psychological phenomena, and what techniques of analysis and modeling might clarify the empirical results?

Much of the empirical work reviewed in this chapter has been conducted within the context of the Western tonal-harmonic style. Across musical styles and cultures, however, the notion of tone (or pitch) centrality can also be found – that is, one central tone anchors a subset of hierarchically related tones. Moreover, it is possible for an individual piece of music to establish its own unique hierarchy. Findings that point to listeners’ sensitivity to such hierarchies are described. Our conceptualiza-tion of the tonal hierarchy therefore invokes a pan-stylistic approach to knowledge acquisition and representation of musical structure.

Psychological research on tonal hierarchies developed beginning in the late 1970s as part of an increasing appreciation of the role of cognition in music. The alternative approach up to that time, with a tradition dating to the ancient Greek

533 A Theory of Tonal Hierarchies in Music

philosophers, focused instead on music acoustics. The basic idea was that the formation of musical structures such as scales and chords could be accounted for by the harmonic structure of complex periodic sounds. The cognitive approach, in contrast, sought to understand the role of experience within the musical culture. It raised a host of interrelated questions, including the psychological processes and neural mechanisms involved in learning musical patterns, the role of development and training, and cross-cultural comparisons. The cognitive approach also encour-aged the development of quantitative models of music learning and perception.

3.2.1 Psychological Principles Underlying Tonal Hierarchies

The structure of tonal hierarchies appears to rely on two basic cognitive principles. The first is the existence of cognitive reference points (Rosch and Mervis 1975; Rosch 1975, 1978, 1979), which motivated the initial empirical studies of tonal hierarchies (Krumhansl 1979; Krumhansl and Shepard 1979; Krumhansl and Kessler 1982). Within categories, certain perceptual and conceptual objects, called cognitive reference points, have special psychological status. They are reference points in relation to which other category members are encoded, described, and remembered. In Rosch’s work they are sometimes referred to as prototypes, although this term seems less apt when applied to musical pitch. Their existence serves the purpose of cognitive economy – that is, an internal coding best suited for making distinctions relevant to the domain in question at the same time conserving finite cognitive resources. Empirical work has been performed on cognitive refer-ence points or prototypes in a wide variety of domains, including visual objects, colors, numbers, faces, and personality descriptions. These investigations have shown that cognitive reference points are given priority in processing, are most stable in memory, and have a special role in linguistic descriptions.

We suggest that not only do cognitive reference points function similarly in music, but also they may be especially important there. This is because music does not provide fixed reference tones except as determined by the music itself. Thus, unlike other domains in which cognitive reference points are defined independently of the category (red is perceptually red whether it is or is not thought of in terms of the category of colors), the function of a tone depends entirely on the musical context. Another way to express this is that for most listeners relational processing (relative pitch) predominates over absolute pitch (with pitches having fixed labels indepen-dent of context). At a general level, the importance of musical reference points is not merely that they exist, but also that they guide musical perception, memory, thought, and understanding.

The second basic cognitive principle is sensitivity to statistical regularities in music. Statistical regularities that have been considered include the distribution of tones (their frequency of occurrence and their total temporal duration), and the frequency of sequences of tones. Recent research has suggested that statistical learning may play a role in language acquisition (Saffran et al. 1996a, b, 1997).


In this research, infants appear to have learned which syllables frequently co-occur in sequences. A learning process such as this may lead to the identification of combinations of syllables as words. Subsequently, the paradigm has been extended to tones (Saffran et al. 1999; Saffran and Griepentrog 2001). Thus, early in develop-ment humans appear to be sensitive to frequent successions of sounds, and this sensitivity may encompass both language and music.

In sum, we propose that regularities within the musical style establish tone centrality. Regularities include repetition of tones and tone sequences, melodic and rhythmic emphasis, durational and metric stress, and positioning of central tones at or near beginnings and endings of phrases. Through repeated exposure to music, listeners implicitly develop a mental representation that captures the regularities. This representation can then be used to encode and remember musical patterns in the future, and generate expectations while listening. Sensitivity to these regulari-ties may also enable listeners to adapt relatively easily to novel musical styles.

3.2.2 Definitions and Distinctions

The concept of the tonal hierarchy draws on a long tradition in music theory and history (DeVoto 1986). Various units of musical structure have been abstracted from compositional practice since the seventeenth century and codified. These include scales, modes, chords, keys, and relations among keys (the circle of fifths), described in basic music texts (e.g., Piston 1987). The notion of a tonal hierarchy incorporates relations among all these units in a stable, abstract frame of reference. In this frame of reference all tones and chords are described with respect to the tone that gives the key its name. For example, in C major the first scale tone is C and it is called the tonic, and the three-tone chord built on it with the tones C-E-G is called the tonic triad. The tone G, which forms a very consonant interval (a fifth) with the tonic, is called the dominant, as is the triad G-B-D that is built on it. Similarly, each other tone and chord is designated relative to the tonic.

The tonal hierarchy does not contain information about pitch height. Octave equivalence is assumed. In other words, all members of a pitch class (e.g., C

1, C

2,

C3, C

4 and so on – where the number refers to the octave containing the tone) are

represented by a single element (in this case, C). Thus, the hierarchy refers to pitch classes rather than to specific pitches. Moreover, the tonal hierarchy does not directly contain information about individual tones as they occur in a musical piece. The order, metric position, timbre, and loudness of tones are not represented. The relative stability of tones in the tonal hierarchy might therefore be characterized as static, independent of the place of the particular tone in the music.

Bharucha (1984) has drawn an important distinction between the tonal hierarchy and the hierarchy created within the framework of a particular piece, or section of music. He named the latter event hierarchy; it describes the relative prominence of events in that particular sequence. “Event hierarchies describe the encoding of specific pieces of music; tonal hierarchies embody our tacit or implicit knowledge of the


abstract musical structure of a culture or genre” (Bharucha 1984, p. 421). So, unlike tonal hierarchies that refer to cognitive representations of the structure of music across different pieces of music in the style, event hierarchies refer to a particular piece of music and the place of each event in that piece.

The two hierarchies occupy complementary roles. In listening to music or music-like experimental materials (melodies and harmonic progressions), the listener responds both to the structure provided by the tonal hierarchy and the structure provided by the event hierarchy. Musical activity involves dynamic patterns of stability and instability to which both the tonal and event hierarchies contribute. Understanding the relations between them and their interaction in processing musical structure is a central issue, not yet extensively studied empirically.

3.3 Empirical Research: The Basic Studies

This section outlines the classic findings that illustrate tonal relationships and the methodologies used to establish these findings.

3.3.1 The Probe Tone Method

Quantification is the first step in empirical studies because it makes possible the kinds of analytic techniques needed to understand complex human behaviors. An experimental method that has been used to quantify the tonal hierarchy is called the probe-tone method (Krumhansl and Shepard 1979). It was based on the observation that if you hear the incomplete ascending C major scale, C-D-E-F-G-A-B, you strongly expect that the next tone will be the high C. It is the next logical tone in the series, proximal to the last tone of the context, B, and it is the tonic of the key. When, in the experiment, incomplete ascending and descending scale contexts were followed by the tone C (the probe tone), listeners rated it highly as to how well it completed the scale (1 = very badly, 7 = very well). Other probe tones, however, also received fairly high ratings, and they were not necessarily those that are close in pitch to the last tone of the context. For example, the more musically trained listeners also gave high ratings to the dominant, G, and the mediant, E, which together with the C form the tonic triad. The tones of the scale received higher ratings than the nonscale tones, C# D# F# G# and A#. Less musically trained listeners were more influenced by how close the probe tone was to the tone sounded most recently at the end of the context, although their ratings also contained some of the tonal hierarchy pattern.

A subsequent study used this method with a variety of contexts at the beginning of the trials (Krumhansl and Kessler 1982). Contexts were chosen because they are clear indicators of a key. They included the scale, the tonic triad chord, and chord


sequences strongly defining major and minor keys. These contexts were followed by all possible probe tones in the 12-tone chromatic scale, which musically trained listeners were instructed to judge in terms of how well they fit with the preceding context in a musical sense. The results for contexts of the same mode (major or minor) were similar when transposed to a common tonic. Also, the results were largely independent of which particular type of context was used (e.g., chord versus chord cadence). Consequently, the rating data were transposed to a common tonic and averaged over the context types. The resulting values are termed standardized key profiles. The values for the major key profile are 6.35, 2.23, 3.48, 2.33, 4.38, 4.09, 2.52, 5.19, 2.39, 3.66, 2.29, 2.88, where the first number corresponds to the mean rating for the tonic of the key, the second to the next of the 12 tones in the chromatic scale, and so on. The values for the minor key context are 6.33, 2.68, 3.52, 5.38, 2.60, 3.53, 2.54, 4.75, 3.98, 2.69, 3.34, 3.17. These are plotted in Fig. 3.1, in which C is assumed to be the tonic. Both major and minor contexts produce clear and musically interpretable hierarchies in the sense that tones are ordered or ranked according to music-theoretic descriptions.

The results of these initial studies suggested that it is possible to obtain quantitative judgments of the degree to which different tones are perceived as stable reference tones in musical contexts. The task appeared to be accessible to listeners who differed considerably in their music training. This was important for further inves-tigations of the responses of listeners without knowledge of specialized vocabular-ies for describing music, or who were unfamiliar with the musical style. Finally, the results in these and many subsequent studies were quite consistent over a variety of task instructions and musical contexts used to induce a sense of key. Quantification

Fig. 3.1 (a) Probe tone ratings for a C major context. (b) Probe tone ratings for a C minor context. Values from Krumhansl and Kessler (1982)


of the tonal hierarchies is an important first step in empirical research but, as seen later, a great deal of research has studied it from a variety of different perspectives.

3.3.2 Converging Evidence

To substantiate any theoretical construct, such as the tonal hierarchy, it is important to have evidence from experiments using different methods. This strategy is known as “converging operations” (Garner et al. 1956). This section describes a number of other experimental measures that show influences of the tonal hierarchy. It has an effect on the degree to which tones are perceived as similar to one another (Krumhansl 1979), such that tones high in the hierarchy are perceived as relatively similar to one another. For example, in the key of C major, C and G are perceived as highly related, whereas C# and G# are perceived as distantly related, even though they are just as far apart objectively (in semitones).

In addition, a pair of tones is heard as more related when the second is more stable in the tonal hierarchy than the first (compared to the reverse order). For example, the tones F#-G are perceived as more related to one another than are G-F# because G is higher in the tonal hierarchy than F#. Similar temporal-order asym-metries also appear in memory studies. For example, F# is more often confused with G than G is confused with F# (Krumhansl 1979). These data reflect the propo-sition that each tone is drawn toward, or expected to resolve to, a tone of greater stability in the tonal hierarchy.

Janata and Reisberg (1988) showed that the tonal hierarchy also influenced reaction time measures in tasks requiring a categorical judgment about a tone’s key membership. For both scale and chord contexts, faster reaction times (in-key/out-of-key) were obtained for tones higher in the hierarchy. In addition, a recency effect was found for the scale context as for the nonmusicians in the original probe tone study (Krumhansl and Shepard 1979).

Miyazaki (1989) found that listeners with absolute pitch named tones highest in tonal hierarchy of C major faster and more accurately than other tones. This is remarkable because it suggests that musical training has a very specific effect on the acquisition of absolute pitch. Most of the early piano repertoire is written in the key of C major and closely related keys. All of these listeners began piano lessons as young as 3–5 years of age, and were believed to have acquired absolute pitch through exposure to piano tones.

The tonal hierarchy also appears in judgments of what tone constitutes a good phrase ending (Palmer and Krumhansl 1987a, b; Boltz 1989a, b). A number of studies show that the tonal hierarchy is one of the factors that influences expectations for melodic continuations (Schmuckler 1989; Krumhansl 1991, 1995b; Cuddy and Lunney 1995; Krumhansl et al. 1999, 2000). Other factors include pitch proximity, interval size, and melodic direction.

The influence of the tonal hierarchy has also been demonstrated in a study of expressive piano performance (Thompson and Cuddy 1997). Expression refers to


the changes in duration and dynamics (loudness) that performers add beyond the notated music. For the harmonized sequences used in their study, the performance was influenced by the tonal hierarchy. Tones that were tonally stable within a key (higher in the tonal hierarchy) tended to be played for longer duration in the melody than those less stable (lower in the tonal hierarchy).

A method used more recently (Aarden 2003, described in Huron 2006) is a reaction-time task in which listeners had to judge whether unfamiliar melodies went up, down, or stayed the same (a tone was repeated). The underlying idea is that reaction times should be faster when the tone conforms to listeners’ expecta-tions. His results confirmed this hypothesis, namely, that reaction times were faster for tones higher in the hierarchy. As described later, his data conformed to a very large statistical analysis he did of melodies in major and minor keys.

Finally, tonal expectations result in event-related potentials (ERPs), changes in electrical potentials measured on the surface of the head (Besson and Faïta 1995; Besson et al. 1998). A larger P300 component, a positive change approximately 300 ms after the final tone, was found when a melody ended with a tone out of the scale of its key than a tone in the scale. This finding was especially true for musicians and familiar melodies, suggesting that learning plays some role in producing the effect; however, the effect was also present in nonmusicians, only to a lesser degree.

This section has cited only a small proportion of the studies that have been conducted on tonal hierarchies. A closely related issue that has also been studied extensively is the existence of, and the effects of, a hierarchy of chords. The choice of the experiments reviewed here was to illustrate the variety of approaches that have been taken. Across the studies, consistent effects were found with many different kinds of experimental materials and methods. Thus, the requirement of converging evidence has been satisfied.

3.3.3 Summarizing the Basic Results: Three Principles of Tonal Hierarchies

This consistency across studies enabled the following theoretical summary to be formulated. Bharucha and Krumhansl (1983; see also Krumhansl, 1990a, pp. 140–152) formalized three principles of tonal stability, the relative position of tones in the tonal hierarchy, as a way of summarizing many of the results just described. They are stated in terms of psychological distance. If two tones are judged as similar to one another then they are said to be separated by a small psychological distance. Or, another measure of similarity is how often they are confused in memory; if there are many instances of confusion between them then they would be said to have a small psychological distance.

The first principle, contextual identity, assumes that not all tones have zero distance from themselves. For example, in a memory task some tones are more often confused


with other tones, whereas others are more often correctly identified as themselves. The principle states that the psychological distance between a tone and itself is smaller (more often remembered and less often confused with other tones) when it is higher in the hierarchy than when it is lower. In the context of C major, for instance, the tone G will be better remembered than the tone F#.

The second principle, contextual distance, states that the average perceived distance between two different tones decreases as their position in the hierarchy increases. For example, all else equal, in a C major context, the tones E and G will be judged as closer than the tones F# and A (because E and G are higher in the hierarchy than F# and A) even though their objective distance (in semitones) is the same.

The third principle, contextual asymmetry, holds that there will be an effect of the order of two tones. When a tone lower in the hierarchy is followed by one higher in the hierarchy they are perceived as psychologically less distant than when the two tones are played in the opposite order. For example, F# will be perceived as closer to G than G is to F#; the same temporal-order asymmetry would be found in instances of memory confusion. Even more specifically, the size of the order difference will depend on difference in the tones’ positions in the tonal hierarchy. For example, the asymmetry between F#-G and G-F# will be larger than the asymmetry between F#-F and F-F# (because G is higher in the tonal hierarchy than F).

These principles were proposed as statements of the psychological effects of the tonal hierarchy independent of the particular experimental measure used, which might be direct judgments, memory accuracy, event-related potentials, or other measures.

3.4 Contemporary Issues that Arise from These Basic Studies

These basic studies have raised a number of issues that are considered next. One issue is whether and how tonal hierarchies are learned. A second is the question of whether tonal hierarchies are musical facts, that is, can be related to objective prop-erties of the music itself. A third is how computational models might serve to understand the structure and origin of tonal hierarchies and how they might use tonal hierarchies to model perceptual processing of music. Finally, we consider the role that tonal hierarchies have played in a recent music theoretic proposal in which it is used to compute distances between musical events and make testable quantita-tive predictions.

Concerning the first issue, a specific learning-based proposal is that tonal hierarchies require extensive experience with music to be internalized. Through repeated and extensive exposure, listeners have learned the relative positions of tones in the tonal hierarchy. Another learning-based approach suggests that learning occurs over a much shorter term. Tonal hierarchies may result from actively process-ing the musical input forming summaries of statistically frequent tones and tone


combinations. If so, then psychological measurements, such as probe tone ratings, may reflect short-term memory for the preceding context. A third nonlearning, psychoacoustic explanation is that the tonal hierarchy reflects acoustic properties of tones. These depend on the harmonics of complex tones in a way that is described later.

To assess these alternatives, a variety of approaches have been taken. Some studies use development and music training as a way to determine the importance of experience on acquiring the tonal hierarchy. Another empirical approach exam-ines individual differences and neurological case studies for abilities allied with the recovery of the tonal hierarchy. This may give clues as to the processes through which tonal hierarchies are acquired. In addition to these commonly used approaches in psychology, music offers another alternative, which is to employ unfamiliar musical styles, for example, from other cultures or nontonal Western music. Another approach is to develop computational models to simulate the empirical results. This approach has the potential for identifying musical features important to establishing tonal hierarchies, and may suggest processes through which they may be established cognitively.

3.4.1 Developmental Studies

The learning-based accounts just outlined assume the tonal hierarchy is internalized through exposure to music. This proposed learning mechanism makes a specific prediction about the developmental course of acquisition. If tonal hierarchies are implicitly acquired through exposure to music, then apprehension and representa-tion of tonal hierarchies will emerge at a later developmental stage than the basic perceptual sensitivities on which they are built. The reason is that if tonal hierarchies are to become internalized as cognitive resources, they require a mature memory system and specialized interactions with environmental resources.

This important statement is associated with a basic conundrum. Because statistical learning may occur in infancy, why does the acquisition of the tonal hierarchy, assumed to be the result of statistical learning, occur relatively late? The problem may be resolved by proposing that although the infant brain has developed to the extent of extracting simple regularities in sound patterns, it has not yet developed those memory resources that along with musical experience allow the extraction of hierarchical regularities among tones.

Numerous studies of infant and child development support this proposal. Regarding basic perceptual sensitivities, during the first year of life infants develop an impressive repertoire; they develop the discrimination of, for example, melodic contours, frequencies, simple harmonic ratios, phrasing, and, to some extent, pitch-scale patterns (for reviews tracing this first year, see Trehub and Trainor 1993; Trehub et al. 1997; Dowling 1999; Cohen 2000; Trehub 2000). However, regarding appre-hension and representation of tonal hierarchies, evidence for appreciation of Western tonal structure does not appear for several more years.


The appearance of a stable tonal center appears to emerge around 5 or 6 years of age. This is the first age at which a stable tonal center is evident in children’s spontaneous singing (Dowling 1999). Similarly, Zenatti (1993) reported a prefer-ence for tonal over atonal contexts at 5 or 6 years (depending on the musical task); the distinction tended to increase over 8–10 years for most children. The ability to process a tonal melody is suggestive of the internalization of a tonal hierarchy that guides encoding and retrieval of melody tones. Trainor and Trehub (1994) evaluated the ability of children and adults to detect changes in a well-structured Western tonal melody. The 5-year-olds in their study were able to detect changes that were out of key but not changes that altered the implied harmony. The 7-year olds and adults noticed changes both to the key and the implied harmony.

An early probe-tone study by Krumhansl and Keil (1982) asked children in first through sixth grade and adult listeners to judge the “goodness” of a six-tone melodic pattern. The initial four tones of the pattern were the major triad tones C-E-C-G presented melodically. The final two tones were probe tones located one octave distant from the triad. The judgments of the youngest children showed only a distinction between scale and nonscale tones. Older children distinguished triad tones from nontriad tones; only the adults isolated the tonic from the other tones of the triad, and revealed the full tonal hierarchy in their judgments. It was noted that for all listeners the scale-step distance between the two probe tones also influenced judgments.

The children’s task was simplified in another study (Cuddy and Badertscher 1987). They asked children in the first to sixth grades to rate one probe tone following the C-E-C-G pattern for “goodness” of completion of the pattern. The pattern and probe tones were Shepard tones (Shepard 1964) intended to reduce attention to pitch distance and focus attention on tonal relations. Under these conditions, Cuddy and Badertscher found that even the youngest children differentiated the levels of the tonal hierarchy – the tonic, the other triad tones, the scale tones, and the nonscale tones. Along with Speer and Meeks (1985), this study also found that major scale patterns were also effective for yielding evidence of the tonal hierarchy.

Subsequent developmental data suggested a subtler trend in the acquisition of knowledge of tone relationships (Lamont and Cross 1994). Collected with two variants of the probe-tone tasks and a game-playing task, their data from a large sample size (N = 285) revealed understanding of scale structure in the youngest children with increasing sophistication up to 11 years of age. Thus, the estimate for acquisition of the tonal hierarchy appears to vary with task demands and strategy, but preschool or early school years are clearly important benchmarks (Shuter-Dyson 1999).

The assumption that tonal hierarchies are implicitly acquired through exposure to music yields the corollary that their apprehension does not require formal music training. Such a claim may seem surprising on first consideration, given the large amount of evidence indicating the advantage of music training for multiple musical tasks. However, the acquisition of the tonal hierarchy may have a privileged status.


That young children reveal this form of tonal knowledge rules out an account based on skills acquired during the study of advanced music theory or required for expert performance.

Thus, for adult listeners, no role of early music training in their representation of the tonal hierarchy would be expected. Here current experimental evidence is somewhat mixed. Some evidence suggests that music training does enhance perfor-mance on tests of tonality (Krumhansl and Shepard 1979; Jordan and Shepard, 1987; Frankland and Cohen 1990; Steinke et al. 1997). Other studies of the tonal hierarchy, however, tend to support the claim that music training may not be required. Relations between probe-tone results and music training were either not statistically significant for the melodic pattern C-E-C-G (Cuddy and Badertscher 1987; Brown et al. 1994) or were slight (Cuddy 2000). Minor differences in proce-dure, such as the choice of pure versus Shepard tones to construct probe-tone stimuli, may be partially responsible for the differences in findings. Untrained listeners tend to focus on pitch height in probe tone responses; as noted earlier, Shepard tones deemphasize pitch height and may help direct the untrained listener to focus on tonal relations. Another difference is sample size, with large samples (N = 100) yielding more statistical power to pick up slight differences (Steinke et al. 1997). Yet even that study showed that the response patterns for all levels of music training were essentially similar to music theoretic descriptions of the tonal hierarchy. The more musically trained simply showed finer distinctions among levels of the hierarchy.

Other paradigms implicating the tonal hierarchy point to the similarity of responses for trained and untrained listeners. In tests primarily designed to assess Narmour’s (1990) bottom-up principles of melodic expectancy, the influence of tonal hierarchy on expectancy judgments was found for both trained and untrained listeners (Cuddy and Lunney 1995; Krumhansl 1995a, b; Thompson et al. 1997). Finally, although music training was associated with improved recognition memory for unfamiliar tunes (Cuddy et al. 1981, 2005), both musically trained and untrained listeners responded similarly to the tonal structure of the tunes.

Taken together, the evidence suggests that a representation of the tonal hierarchy exploits general perceptual predispositions, is acquired in childhood through accul-turation, and is acquired without formal intervention. Tonal melodies are easier to recognize than nontonal melodies; pitch alterations in melodies are easier to detect if the alteration deviates from tonal rules than if it does not. Music training is not needed to facilitate acquisition of the tonal hierarchy but, rather, teaches skills and strategies to apply this knowledge to musical problems. (For supporting arguments, see Bigand and Poulin-Charronnat 2006; Marmel and Tillmann 2009.)

Moreover, this section has elaborated an application of the second basic cogni-tive principle underlying tonal hierarchies, that of sensitivity to statistical regulari-ties in music. This sensitivity is not dependent on music training, but is dependent on the maturation of a memory system capable of dealing with tonal/hierarchical regularities. Thus musical memory and representation of the tonal hierarchy are intimately associated. Across individuals, covariation between these two compo-nents of musical cognition may be expected.


3.4.2 Individual Differences and Neurological Case Studies

The argument is that the tonal hierarchy provides a stable framework for accurately encoding and remembering musical pitch patterns. From this argument it follows that the cognitive representation of the tonal hierarchy should be associated with musical memory. One test of this statement is to examine individual differences in associations between tonal hierarchy and musical memory. Individuals differ in perceptual and cognitive abilities and also in early musical environments. Thus, individual differences should be revealed as follows: the ability to recover the tonal hierarchy is reflected in proficiency at musical memory; failure or loss of the tonal hierarchy is accompanied by musical memory failures.

Striking observations of individual differences result from the comparison of neurologically disordered individuals and healthy age-matched controls (see, e.g., the special April 2008 issue of Music Perception which is devoted to Music and Neurological Disorders). Neurological case studies examine patterns of loss and sparing of abilities, called “dissociations.” One line of evidence has been obtained from case studies of individuals with localized brain injury resulting in selective loss of musical abilities (for reviews, see Peretz 1993a; Dalla Bella and Peretz 1999; Marin and Perry 1999).

Interpreting the early historical evidence is a challenge. The evidence is anec-dotal, not collected under controlled conditions. Individual patients may display complex, not regular, patterns of symptoms and many questions regarding cerebral localization of musical function remain unanswered. Nevertheless, recent case studies of amusia – clinical disorders of music abilities due to brain damage – have reported data collected under controlled conditions and reveal much about the func-tional architecture of the brain.

Peretz and colleagues (e.g., Peretz 1993b, 1996; Peretz et al. 1994, 1997; Liégeois-Chauvel et al. 1998) have systematically investigated three patients, CN, GL, and IR, who sustained musical deficits after surgery involving bilateral damage to auditory cortex. An extensive program of testing included music, speech, language, and other cognitive skills. Case results were compared with results from controls matched for age, education, and music background (all were nonmusicians). A fourth patient, KB, an amateur musician who suffered right frontoparietal damage after a stroke, was similarly evaluated by Steinke et al. (2001; see also Lantz et al. 2003).

All four patients had normal language and intellectual functioning and all four failed various tests of musical functioning. Thus, the results clearly demonstrate dissociation (separation) of music and language in brain organization. Within the music tests, the patients had difficulty with both tests of tonality and tests of recog-nition memory for familiar music (for the latter, KB’s difficulties extended to instrumental, but not song, tune recognition). The connection between representa-tion of the tonal hierarchy and musical memory is supported most clearly in the data of CN and GL. These two patients, unlike IR and KB, showed good recovery (CN) and sparing (GL) of perceptual skills. Thus, failures at the tests of tonality and memory are not likely due to impaired perception of tones.


For example, 6 years post-onset CN performed at a normal level on auditory perceptual tasks, such as discrimination of isolated musical pitches, detection of contour and interval (frequency ratio) changes in novel musical sequences, and detection of rhythmic changes in novel musical sequences (Peretz 1996). However, CN was still markedly impaired at recognizing and naming familiar melodies, and classifying melodies as familiar or unfamiliar. She was also impaired at rating probe tones with a tonal melody as musical context, at judging the appropriate tonal ending for melodies, and at remembering the pitches of unrelated tones (Steinke et al. 1994, 1997). Given a probe-tone task (that of Cuddy and Badertscher 1987), GL’s responses showed some evidence of contour and interval processing, but no sensitivity to tonal function (Peretz 1993a, b). GL was also deficient at melody recognition and pitch memory. Peretz (1996) suggested that tonal encoding of pitch (coding in terms of tonal function) is a major determinant of access to stored musical representations. From a review of the evidence, it is probably critical. Impairment of acquired cognitive references and/or their implementation leads to difficulties with the probe-tone and other tonality tests and severe failures of melody recognition and memory.

Complementary evidence also supports a link between tonal encoding and musical memory. This evidence has been obtained in case studies of (probable) Alzheimer’s disease (AD) – for example, Cuddy and Duffin (2005), Fornazzari et al. (2006), and Vanstone and Cuddy (2010). Rather than displaying musical impairments after stroke despite cognitive recovery, some AD individuals show the reverse pattern of dissociation. They demonstrate preserved musical memory despite severe speech and other cognitive difficulties. Of importance, in the studies by Cuddy and colleagues, is that these persons detected tonal errors in tunes as accurately as did age-matched healthy controls. Thus intact tonal encoding of pitch may have facilitated access to stored representation of tunes.

As a final example under the topic of individual differences, cases of congenital amusia may be noted. Unlike brain damage, congenital amusia (or “tone deafness”) is not acquired through injury or assault. It is considered a developmental disorder (e.g., Ayotte et al. 2002; Hyde and Peretz 2004; Peretz et al. 2008) and may be a neurogenetic anomaly (Drayna et al. 2001; Peretz 2008). The auditory skills of congenital amusics are neurologically normal, with the outstanding exception of the inability to develop normal musical abilities such as recognizing a familiar tune in the absence of lyrics and detecting out-of-key notes in conventional melodies. They lack “the (implicit) knowledge and procedures required for mapping pitches onto musical scales” (Peretz et al. 2008, p. 332).

Thus, in certain cases of acquired and congenital amusia, both musical memory and tonal representations are compromised. Musical memory, it has been sug-gested, depends on top-down activation of a cognitive framework for encoding and storing musical materials. As revealed by studies of individual differences and neurological case studies, the tonal hierarchy is a crucial component of this framework.


3.4.3 Tonal Hierarchies and Tone Distributions in Western Music

If listeners acquire the tonal hierarchy by internalizing statistical regularities on the musical surface, then psychological measures of tonal hierarchies should correlate with tone distributions in musical compositions. This section reviews a number of studies that find the tonal hierarchy measured in empirical research mirrors the emphasis given the tones in compositions by frequency of occurrence and duration. This relationship between subjective and objective properties of music provides a strong musical foundation for the psychological construct of the tonal hierarchy.

According to Meyer (1956), musical styles consist of systems of probability relationships that capture the characteristic patterns in the style. He goes on to say that the meaning of any given musical event depends on how it enters into the pat-terns captured in the probabilities. Thus, the hierarchy evident in the probe-tone ratings, and the variety of other behavioral measures summarized earlier, might be related to the statistical distributions of tones in music.

This relationship was first examined by Krumhansl (1985; see also Krumhansl, 1990a), who compared the tonal hierarchies with various statistical treatments (Youngblood 1958; Hughes 1977; Knopoff and Hutchinson 1983). These statistical treatments tabulated the frequency of occurrence or total duration of each tone of the chromatic scale in pieces by Schubert, Mozart, Hasse, Strauss, Mendelssohn, and Schumann. Tone distributions correlated strongly with the probe-tone profile for the corresponding key; tones high in the tonal hierarchy tend to be sounded more frequently and for longer durations. Similar results were found by Järvinen (1995) in jazz improvisations, especially at more stressed metrical positions. Minor discrepancies between the statistics and the probe-tone profiles have been noted (Krumhansl 1990a, p. 69; Auhagen and Vos 2000) but these were relatively minor and can be explained in large part by proximity to the tonic.

Huron has been at the forefront of analyzing music for statistical properties (see the summary in Huron 2006). He has also developed and distributed tools for statistical studies of music (Humdrum Toolkit and Themefinder.com). He stresses the adaptive value in evolution of being able to anticipate frequently occurring events (Huron 2006, p. 357). As a general cognitive principle it is adaptive because knowing what, when, and where something is likely to occur speeds perception, action, and evaluating the consequence of alternative actions. He also notes the advantages of statistical learning, “In a stable environment, the most frequently occurring events of the past are the most likely events to occur in the future. Thus, a simple yet optimum interactive strategy is to expect the most frequent past event.” (p. 360, italics in the original)

In an early study, Huron (1993) analyzed a large sample of Bach’s music. He looked at which tones in chords are most frequently sounded in two different octaves (doubled). The number of these doubled tones correlates strongly with the


probe-tone profiles. He concluded that the doubling of the tones high in the tonal hierarchy reinforces the perception of key. Huron’s collaborator Aarden (2003) conducted the largest note count to date. As shown in Huron (2006, pp. 148–149), Aarden found the distribution of tones in music (when modulating passages were excluded), with more than 65,000 notes for melodies in major keys and more than 25,000 notes for melodies in minor keys. Aarden’s note counts differ somewhat from previous ones. This may reflect the different compositions analyzed, but Aarden went on to show that listeners’ expectations conformed better to his statisti-cal results than to previous ones. Thus, the behavioral measures were accounted for better by his musical analysis of tone distributions than by other analyses.

Although the details of the tone distributions in these studies differ in detail, they all point to a common conclusion. Subjective properties of music as studied in psychological research correlate well with objective properties of music, specifi-cally the relative emphasis to the tones by frequency and duration. The distribu-tional emphasis would serve the purpose of initially establishing and then maintaining the listener’s sense of the tonal reference points. The extent to which the probe-tone ratings resemble the tone distributions supports Meyer’s proposal that listeners have internalized the statistical properties of music.

3.4.4 Cross-Cultural Studies of Tonal Hierarchies

One way to study the acquisition of tonal hierarchies is to use unfamiliar styles from non-Western cultures. It is important to look at tonal hierarchies cross-cultur-ally to explore their generality across styles. Moreover, it is important to rule out competing explanations of their psychological basis. That is, the results just dis-cussed for Western music may be a function of some other musical attribute. For example, perhaps consonance influences both the distribution of tones and the measured tonal hierarchy. More specifically, Western music is organized around harmony (chords) and so it might favor tones that form consonant intervals with the tonic and other tones that are consonant with the tonic, and this might determine the tonal hierarchy. This can be addressed by studying music that is not organized around Western harmonic structure.

Two studies bear directly on the acquisition of tonal hierarchies. These collected probe-tone ratings with contexts drawn from non-Western music. Castellano et al. (1984) used contexts from 10 North Indian ragas. One of the most significant dif-ferences from Western music is that the primary means of expressing tonality in Indian music is through melody. In addition, North Indian music has a greatly expanded set of scales (called thats, all with the same tonic) compared to the major/minor system of Western music. Theoretical treatments of Indian music describe a hierarchy of the importance of tones. In a probe tone study, ratings of the 12 tones of the Indian scale largely confirmed these predictions.

It was surprising that the results for both Western and Indian listeners agreed with the theoretical predictions for the Indian music. This was surprising because


it indicates that the Western listeners adapted quite readily to tonal hierarchies in this unfamiliar style, rather than requiring extensive experience. Both groups gave high ratings to the first (Sa) and fifth (Pa) scale tones, which are described as most structurally significant and which are sounded in the drone accompany-ing the melody. They also gave relatively high ratings to the vadi tone, which is a tone given emphasis in the melody and is specific to each raga. To explain the agreement between the two groups of listeners, the contexts were examined. It was found that the theoretically important tones were sounded more frequently than other tones. Apparently, the unfamiliar listeners used this information about the distributions of tones to make their judgments. It suggests that distri-butions of tones can convey the tonal hierarchy to listeners unfamiliar with the style. Beyond this level of agreement, the Indian listeners were more sensitive to scale (that) membership. The important finding, however, is that the results demonstrate a strong link between objective and subjective measures of tonal hierarchies.

Similar results were found in an experiment with the music of Bali (Kessler et al. 1984). The study of Kessler et al. had an expanded design compared to the study with North Indian music. It used contexts from both Western and Balinese melo-dies, and both Western and Balinese listeners some of whom were unfamiliar with Western music. Balinese (and Javanese) music is interesting because it uses two different tuning systems, sléndro and pélog, both of which are different from Western (diatonic or chromatic) tuning (see also Perlman and Krumhansl 1996). Their study included both sléndro and pélog contexts. Listeners’ responses revealed a number of strategies. Some, but only some, of the Balinese subjects produced results corresponding to the predicted tonal hierarchies. Other Balinese subjects responded primarily to pitch height, giving higher ratings to higher tones. Western listeners responded to frequency of occurrence of the tones in the context, which also correlated to some degree with the predicted hierarchy. Again, this suggests that naïve listeners can use tone distributions to perceive the major anchoring tones of the tonality.

To our knowledge these two studies are the most directly related to tonal hierarchies in particular. Two other studies demonstrate in cross-cultural studies that the tonal hierarchy is one factor entering into judgments of melodic continua-tions (Krumhansl et al. 1999, 2000). The first of these used Finnish folk hymns that are passed on by oral tradition by conservative religious sects in southwest Finland. They combine elements of Finnish folk music and Lutheran hymns as they came to be known in Finland in the early eighteenth century. They contain modes and other features unlike tonal–harmonic music. The second study used Sami yoiks, which are also purely oral and improvised around short, repeated motives. Most are based on the five-tone pentatonic scale and the intervals in the yoiks tend to be larger than those found in most Western melodies. Listeners unfamiliar with the style demon-strated that they were able to orient to the appropriate tonal hierarchy based on the relative emphasis given the tones in the music. As in other studies, the melodic continuation judgments of listeners familiar with the style exhibited style knowl-edge over and above the salience of tones in the context.


Recent research in music cognition has expanded to include styles from many different cultures, studying not only the way in which the music is perceived and understood, but also its emotional effects. Cross-cultural studies raise a vast number of interesting questions and may reveal underlying cognitive principles not yet identified in music research.

3.4.5 Tonal Hierarchies in Nontonal Western Music

More recent Western music has introduced styles of music that depart from tradi-tional tonal structure. In other words, they are not structured around the system of major and minor harmony that dominates Western music. These novel styles raise the question as to whether the structural principles with which they are written can be perceived and understood by listeners. Twelve-tone serialism has been particu-larly controversial. In addition to investigating the listener’s response to the styles generally, these studies provide additional materials with which tonal hierarchies can be examined.

Two studies used the probe-tone method with bitonal contexts, in which tones from two different keys are played simultaneously. The style raises the question as to whether listeners can process two keys as separate entities or whether the resulting perception is one of a fused combination of the two. In the first of these (Krumhansl and Schmuckler 1986), the context was the Petroushka chord from Stravinsky’s ballet. It is a striking example of bitonality that uses the tonic triads of the keys of C and F# major. These keys have maximally dissimilar tonal hierarchies (Krumhansl and Kessler 1982). This would seem to optimize the possibility that the materials from the two keys could be heard independently, as there would be little confusion as to which key a tone belonged. Indeed, the probe tone ratings showed the influence of the two keys, but two experiments with selective attention tasks (wherein participants were instructed to focus on just one key) established that listeners cannot perceptually segregate the materials from the two keys. This was true even for a group of musicians who had recently performed the piece in concert, including the two clarinetists playing the part. These findings suggest that the two component keys are not perceptually functional as independent entities. Rather, the context appears to establish a composite hierarchy of the tones of the two triads.

The second study of bitonality by Thompson and Mor (1992) found more positive evidence for listeners’ abilities to process two keys at once. Their two excerpts (from pieces by Dubois and Milhaud) contained materials from the keys of C# and F major, one represented in the upper stave and the other in the lower. In both, the materials in the two staves were quite distinct. In the first, the music in the upper stave was melodic whereas that in the lower stave was chordal. In the second, the music in the upper stave was considerably higher in pitch than that in the lower stave. With these differences, the listeners’ responses reflected long-term knowl-edge of tonal hierarchies, not just the distribution of tones in the contexts. In the


case of the first except, evidence for two functional keys was found; in the case of the second excerpt, the tonality of the material in the upper voice strongly predominated.

Krumhansl et al. (1987) examined another unconventional style, 12-tone serialism. Twelve-tone serialism is an influential style in which many compositions were written in the twentieth century. It is of special interest here because it is specifi-cally intended to oppose traditional ways of structuring music in terms of chords and keys, not allowing a hierarchy of salience to emerge. It does this by requiring that all tones of the chromatic scale be sounded before any of them is repeated. The ordered series of the 12 chromatic tones is called the series, or tone row. This means that no tone receives particular emphasis. This study differed from those reviewed in the last section because it found different results depending on familiarity with the style.

The experiments used materials from two compositions by Schoenberg (Wind Quintet and String Quartet, No. 4), who is generally considered the innovator of the style. Two groups of listeners participated, both of which were musically trained. However, only one group was familiar with the 12-tone style primarily through academic study. Their probe tone ratings showed evidence of reversing the normal pattern for major and minor tone profiles – that is, they gave high ratings to tones that denied local implications of key. This is consistent with the intention of the style to avoid tonal implications. In contrast, the probe tone ratings of the group of listeners unfamiliar with the style showed influences of local tonal implications, contrary to the style’s intention, and also gave higher ratings to tones sounded more frequently, as in other studies reviewed here, and also more recently, which would be expected from short-term memory. Parenthetically, neither group of listeners seemed to have internalized the ordered sequence of tones in the series. They did not give higher ratings to tones following the contexts in the series, despite rather extensive experience with the series in the course of the experiments.

In sum, these studies of nontonal Western music show that the influences of major and minor tonal hierarchies are more complex than in traditional Western music. The experimental results are in some cases consistent with the composi-tional methods, but in other cases they are not. Given the vast number of recent stylistic innovations, these studies can be viewed only as a minor inroad into the interesting questions they pose.

3.4.6 Novel Tone Sets

The work cited in the preceding text indicates that the distribution of tones in composed music supports the tonal hierarchy. Frequencies of tone occurrence and tone duration both tend to correspond with position in the tonal hierarchy. According to Krumhansl (1990b), “the primary significance of the observed correspondence is to suggest a mechanism through which the principles of musical organization are learned” (p. 315). With repeated listening to music, the consistent features of the


tone distribution become internalized as an abstract internal schema or framework (Krumhansl 1987, 1990a).

The question then arises as to whether the internal schema, thus acquired, interferes with or even cancels the listener’s ability to pick up pitch distributional information in a novel idiom – that is, a distribution not convergent with the Western tonal hierarchy. If the pitch distribution information in a musical context is not conver-gent with the tonal hierarchy, one of several possibilities may result. The listener may be unable to pick up or remember conflicting information; the information is either not processed or not retained. Alternatively, a listener may attempt to assimi-late the conflicting information to a more familiar tonal hierarchy (Dowling 1978). Perceptual judgments may be systematically distorted with respect to the pitch distribution (Jordan and Shepard 1987). Or, finally, listeners’ strategies for the abstraction of pitch structure may be flexible and adaptable. Thus, they may easily abstract distribution information that deviates from the conventions of the Western tonal idiom (Krumhansl 1990a).

Oram and Cuddy (1995) addressed the question in the following way. They constructed melodic sequences of 20 pure tones, each tone of duration 200 ms. Each sequence was generated from either a diatonic tone set, C, D, E, F, G, A, B, or a nondiatonic tone set. If nondiatonic, the tone set did not conform to the scale of any major or minor key. Within each sequence, the frequency of occurrence of each tone was determined according to the following ratios: one tone of the tone set occurred eight times, two tones occurred four times each, and the remaining four tones occurred just once. The tones selected to occur most frequently never formed the simple pattern of a major triad in root position. Both musically trained and untrained listeners heard the sequences in a probe-tone paradigm. The results for both levels of music training were quite straightforward. Probe tone ratings were systematically related to the frequency of occurrence of the tones in the sequence.

Musical knowledge, however, did play a role. First, the effect of frequency of occurrence within the melodic sequence was more pronounced for musically trained than untrained listeners. This finding is surprising and somewhat counterintui-tive. Yet it is both consistent with the notion of flexibility of musical pitch processing and suggestive that music training does not tie processing to a rigid schema – rather it fosters a strategy for the pick-up of novel distributional information.

Second, the effect of frequency of occurrence was more pronounced for diatonic sequences than for nondiatonic sequences. This indicates that prior knowledge of diatonic structure can be coordinated with sensitivity to tone distributions. Third, the data for the musically trained listeners revealed some degree of assimilation to the Western tonal hierarchy for diatonic sequences. Along with frequency of occur-rence and pitch proximity (pitch distance of the probe tone from the last tone of the sequence), relative stability in the tonal hierarchy added to prediction of the probe-tone ratings. The important conclusion from this study, however, is that listeners are flexible and adaptable to novel pitch distributions, and that extensive music training does not appear to interfere with this sensitivity – if anything, it appears to enhance it.

Cuddy (1997) reported a follow-up probe-tone study with composed musical melodies. As before, sequences were generated from either a seven-tone diatonic or


nondiatonic tone set, but the sequences were now 20-s original flute compositions. The total duration of one tone, summing across each occurrence of the tone, was 8 s; of two different tones, 4 s, and of four different tones, 1 s. Within the total duration allotted to each tone, composers were allowed to distribute the duration of each occurrence of that tone and its octave location according to their own inten-tions or style. Despite differences in musical materials, the results closely replicated those of Oram and Cuddy (1995). Durational differences in the musical surface were reflected in probe tone ratings.

Cuddy (1997) also reported data from a second study in which a pair of probe tones followed each presentation of the flute melody. The listener was asked to judge how related the first tone was to the second tone of the pair with respect to the melody. For Western tonal harmonic contexts, relatedness is judged by order of the tonal stability of the two probes, as noted earlier in this chapter (Krumhansl 1979; 1990a; Bharucha and Krumhansl 1983). With the novel flute melodies, on the other hand, judgments of relatedness were predictable not from tonal stability, but from the duration biases of the melody. They were higher for the order in which a probe tone of shorter total duration in the melody was followed by a probe tone of longer duration in the melody, than the reverse.

This section proposed that the correspondence between distributional properties in music and the tonal hierarchy is important because it suggests a mechanism for how the tonal hierarchy is acquired. Work with novel tone sets reveals that listeners do possess a finely tuned ability to discover distributional regularities, an ability that is a necessary prerequisite to learning. They also preserve the ability to orga-nize tones in new musical contexts.

3.4.7 Tone Distributions: Frequency or Duration?

A question that arises from these studies is whether pitch salience depends on the frequency or duration of the tones in the inducing context. The paradigm with novel tone sets allows researchers to tease apart the role of distributional cues that are tightly coexistent in music. The mechanisms for the cognitive processing of fre-quency of occurrence – as reflected by frequency of tone onset – may differ from those for processing tone duration. The (controversial) argument is that the two processes are separable both at the neural level (Whitfield and Evans 1965) and at the general level of cognitive principles (Yonelinas et al. 1992). It is thus instructive to consider whether separation is applicable to musical contexts. If it is, implica-tions follow for the relative importance of cues in key-finding both by listeners and by computational models (see the next sections).

Salience was manipulated in Oram and Cuddy (1995) as the frequency of occur-rence of isochronous tones; thus frequency of occurrence and total duration covaried. Salience in Cuddy (1997) was defined as total duration of tones with frequency and duration of individual events free to vary. Nevertheless, the findings were the same: Tones of greater surface salience in the sequence, not the diatonic tonal hierarchy,


acted as reference points or anchors for other tones of lesser surface salience. Of the possible outcomes outlined at the beginning of this section, the one favored is that listeners’ strategies for the abstraction of pitch structure were flexible and adaptable. Representation of the Western tonal hierarchy did not interfere with the abstraction of distribution information that deviated from familiar convention.

Lantz and Cuddy (1998) constructed nontonal melodic sequences in which frequency of occurrence and duration were pitted against each other. For example, in a given sequence, certain tones occurred more frequently, while the duration (note values) of other tones was longer. In addition, the total duration of the longer tones was varied. Using a standard probe-tone paradigm, the researchers found that tones sounded with longer duration, and tones sounded for greater total duration, were rated higher than other tones. Frequency of occurrence per se did not influ-ence listeners’ ratings under these conditions.

In related work, Smith and Schmuckler (2004) also used the probe-tone technique to examine cues of frequency of occurrence and duration. Their technique of sequence construction was slightly different from those of the aforementioned studies. Their sequences contained all 12 tones of the chromatic scale presented in random order. The assignment of frequency of occurrence and duration to each of the 12 tones was varied across experiments. Moreover, the duration of each tone was assigned to correspond or not to correspond with the salience of each tone in a given key of the Western tonal hierarchy. Convergent with the aforementioned studies, Smith and Schmuckler (2004) found that duration was a more salient cue to structure than frequency of occurrence. They also noted that duration was increasingly effective as the absolute difference increased between the duration values assigned to tones. However, their listeners were sensitive to duration only when both the individual and total duration of tones corresponded to the Western tonal hierarchy. In other words, listeners’ ratings did not differentiate reliably among tones if the assignment of duration values did not correspond to this hierarchy. This work, therefore, suggests some limitations on listeners’ sensitivities to novel pitch distributions. One possible factor playing a role in setting limitations is the overall pitch information load. Listeners’ capacities for apprehending novel pitch distributions through duration cues may be limited to tone sets of six or seven tones (as is found in most musical scales) and may not extend to tone sets of 12 tones.

In sum, research with novel tone sets promises to uncover sensitivity to the cues leading to acquisition of the tonal hierarchy and also sensitivity to cues pointing to nontraditional or nontonal hierarchies. More research is invited to clarify the extent and boundaries of these sensitivities.

3.4.8 Tonal Hierarchies and Computational Models

Many computational models of tonality have been proposed and tested against psychological data. Some of them focus specifically on the tonal hierarchy. One reason for this is that the quantified tonal hierarchy provides detailed data against


which the computational models can be tested. Different models make different assumptions about the processes and levels of processing at which the tonal hierarchy is generated and the way it is manifested in the output of the model.

One model uses psychoacoustics, that is, low-level processes to generate tonal hierarchies. This model tests the extent to which psychoacoustic processes can account for tonal hierarchies without assuming cognitive processes. At the core of Parncutt’s model (1989; 1994; Huron and Parncutt 1993) is the notion of pitch salience, which derives from Terhardt’s (1979; Terhardt et al. 1982a, b) model of pitch perception. This approach posits subsidiary pitches that are not physically present, but that arise through the interaction of frequencies that are present. These are called virtual pitches. The virtual pitches are weighted according to how well the spectral components match its harmonic series or harmonic “template.” In one application, Parncutt (1994) found that this approach could model quite well the tonal hierarchies of some of the context sequences used in the first experiment of Krumhansl and Kessler (1982). The correlations with the probe-tone results were quite high. To eliminate the possibility that the probe-tone profiles simply reflect the tone distributions in the experimental contexts, they also compared the results of their model with the distributions of tones in the contexts. That these correlations were lower argues against the idea that the probe-tone profiles simply reflect tone distributions in the contexts, and suggests that the virtual pitch approach may account for some of the patterns in the probe-tone profiles that go beyond the presence or absence of tones. In an extension of this model, Huron and Parncutt (1993) combined the pitch salience approach with sensory memory decay and again found fairly close results.

Another model uses a multilevel approach, assuming processes at both psychoa-coustic and cognitive levels. Leman’s (2000) model incorporates both virtual pitches, as in Parncutt’s model, and short-term memory. The input to the model is acoustic, which is processed by a peripheral auditory model (simulating the filtering of the ear), and then analyzed for periodicity pitches. The summed “completion image” is similar to the virtual pitch model just described. This pitch module is then entered into an echoic memory module, which incorporates both integration and decay over time. Applied to the chord contexts of Krumhansl and Kessler (1982), the model produced tonal hierarchies that are similar to the probe-tone profiles. From this, Leman concluded that the probe-tone profiles could be accounted for by short-term memory for the perceptually immediate context. There are a number of problems with this conclusion, the most decisive of which is that his input was the composite of all harmonic contexts in Krumhansl and Kessler (1982), not the indi-vidual harmonic contexts. Direct comparisons would show marked discrepancies. Moreover, this model’s result for scale contexts primarily reflects scale member-ship and not the probe-tone data.

A different, more dynamic, approach involves self-organizing neural networks. Such a model has been presented by Tillmann et al. (2000). A self-organizing map creates a topographic mapping between the input data and neural net units on the map. Before learning, the map units have no particular organization, but when presented with input data an ordering appears over time. The model contains three layers,


corresponding to tones (pitch classes), chords, and keys (only major keys were represented). Different training sets were used: simple (idealized) harmonic sequences and more realistic harmonic sequences. These were either “sparsely coded” (just the presence of the tones was coded) or “richly coded” (including psychoacoustic pitch salience following the scheme of Parncutt 1988). In other words, one model assumes that the pitches are simply perceived as such, and the other model takes into account psychoacoustic properties.

After training, the neural network was tested against various previous experi-mental studies. In one test, their trained model was “presented” with major and minor chords. The resulting activation of tone units resembled the probe-tone profiles of Krumhansl and Kessler’s (1982) first experiment. Their model was also com-pared with the results for the three major key sequences of Krumhansl and Kessler’s (1982) second experiment (the model only included major keys). In that study, sequences of chords were played stopping at each successive chord for probe tone judgments in order to trace how the sense of key developed and changed over time. Their model traced quite closely these changes in the sense of key. Finally, the neural network model reproduced the direction and magnitude of similarity judg-ments for sequentially presented tones. (Recall that two tones are judged to be more related if the second tone is higher in the tonal hierarchy than the first, compared with the opposite order; Krumhansl 1979, 1990a).

Together, the models described in this section offer suggestions about possible psychological mechanisms that contribute to the perception of the tonal hierarchy. They variously include virtual pitches created by interactions of physically present harmonics, memory traces of the perceptually immediate context, and the result of internalizing regularities from harmonic sequences over an extended period of time. At present, no single account seems entirely adequate to account for the wide range of behavioral data currently available on tonal hierarchies. It seems that the most likely outcome of further modeling efforts will be that the behavioral data rely to some degree on all three of these psychological mechanisms, as well as others, and that these various mechanisms are not clearly independent of one another.

3.4.9 Using Tonal Hierarchies in Key-Finding Models

A number of models have addressed the problem of key-finding. Key-finding refers to the process through which a listener initially orients to the key of a piece of music and subsequently reorients to new keys if modulations (changes of key) occur in the music. It is important for music perception because the function of tones in melodies and chords depends on their relationship to the tonic, as described previously.

The objective of key-finding models is to take some musical input and assign a key (or keys) to it. Models vary as to whether the input is acoustic information or symbolically coded music (such as score notation or MIDI code). The choice of input also depends on the objective of the model; it may be an entire piece of music,


a short segment from the initial or final portion, or a sliding window of some length. Some models allow retrospective reevaluation of key assignment. Some models assign a single key and other models allow for the possibility that a number of keys might be quite strongly suggested (or no key may be perceived as very strong). Key-finding models may or may not attempt to characterize the process through which the listener finds the key, however. That is, some attempt to be psychologically realistic, others draw on music theory, and others are shaped by computational con-siderations. Whether intended as a psychological model or not, automatic determi-nation of key has utility in applications, for example, determining the key is prerequisite to successful automation of music analysis (Rowe 2000). Considerable modeling effort has been devoted to the problem of how to characterize this process (cf. Vos and Leman 2000), and only those that use tonal hierarchies are described here.

Likely the first key-finding algorithm to be implemented on a computer was that of Longuet-Higgins and Steedman (1971). This algorithm uses a two-dimensional array of tones. It matches incoming tones of a piece of music to box-shaped regions containing the scale tones of a key. In their model there is one region for each major and minor key (the harmonic minor scale is used). The algorithm works by elimi-nating musical keys as the music progresses. Special rules are applied if the model finds either no or multiple keys at the end of the process. Its results can be com-pared with models that assume that over and above scale membership the tones are differentiated more finely, as in the tonal hierarchies.

This idea motivated the development of a different key-finding algorithm that weights incoming tones according to the tonal hierarchy. Krumhansl and Schmuckler (Krumhansl 1990a) suggested this approach might result in a more accurate and efficient algorithm than simply assessing whether or not tones are scale members, as was done in the Longuet-Higgins and Steedman model. The input to the algo-rithm is the distribution of tones in the input segment weighted according to their duration. That is, it is based on the summation of tone durations (total duration) of each of the chromatic scale tones in the segment. The algorithm correlates this input with the 24 major and minor Krumhansl and Kessler (1982) key profiles. The cor-relations give a measure of the strength of each possible key. The algorithm was compared with that of Longuet-Higgins and Steedman (1971), with favorable results, and Cohen’s (1991) listeners’ judgments of key for the Bach Preludes and Fugues in the Well-Tempered Clavier. To visualize the results, they were projected onto the geometric map of keys that resulted from correlating the tonal hierarchies of different keys (Krumhansl and Kessler 1982; Krumhansl 1990a).

This led to the development of a dynamic version of the model using a self-organizing neural net model (Krumhansl and Toiviainen 2001; Toiviainen and Krumhansl 2003). It used data collected using a continuous probe-tone task in which the probe tone was sounded continuously while the music was played. Listeners heard the entire piece of music (a Bach organ piece) and rated how well one probe tone fit with the music as it progressed in time. Then another probe tone was chosen and the procedure was repeated until all 12 chromatic scale tones were used. This produces a probe-tone profile at each point in time during the music. These results were projected onto the self-organizing neural net resulting from


training it on the probe-tone results of Krumhansl and Kessler (1982), which replicated the earlier geometric map of keys. The perceived tonality was repre-sented as continuously changing patterns of color on the self-organizing net, with color representing the strength of the keys. This was compared to an alternative model using just the tone distributions in the music. The listeners’ data showed influences of tonality over and above the emphasis given tones by frequency and duration in the music reflecting knowledge of the style beyond tone distributions.

A major contribution in key-finding models comes from the work of Temperley (1999; 2001; 2007; Temperley and Marvin 2008). Temperley (1999) suggested a number of modifications of the Krumhansl-Schmuckler algorithm. One modifica-tion, based on music-theoretic considerations, was to modify the weights of the tones. Another is that a penalty was imposed for changing key from one input segment to the next. Finally, a retrospective reevaluation of key was permitted. In addition to comparing the model with that of Longuet-Higgins and Steedman (1971), Temperley carried out the most extensive test to date using a music theory textbook’s analysis of key in a large number of pieces. This was done on a measure-by-measure basis. The modified model performed well in these applications. Temperley (2001) describes additional theoretical concerns, model modifications, and tests of the model. It should be noted that more recently Temperley (2007) introduced an alternative approach to key-finding that is based on the probability of sequences of tones in different keys and assigns key according to these probabilities (using Bayes’ theorem from statistics).

The models reviewed here utilize tonal hierarchies for finding the key of a musical selection, where the selection can range from a few notes to entire sections of music. The models to date indicate the utility of this modeling approach in these applications. Once the key is identified, then the functions of the tones in the key are determined, allowing subsequent processing, such as analyzing the music for the harmony (chords).

3.4.10 Tonal Hierarchies and Musical Tension

As one listens to music, there are points in time, such as at phrase endings, where the music seems relatively complete and the feeling of tension is low. At other points in time, the music generates strong expectations that the music must con-tinue onto some resolution, with high levels of tension. The rising and falling of tension is fundamental to the experience of music and may contribute to the emo-tional response (Meyer 1956; Krumhansl 2002; Krumhansl and Agres 2008). The question addressed in this final section is how a model might account for these variations in tension over time.

Lerdahl’s (1988, 1996, 2001, 2009) pitch space model makes quantitative predictions of the how the degree of tension varies over time. The music is first analyzed as chords in keys (the standard harmonic analysis, such as a G major chord in the key of C major). The music analyzed in this way, called events, is the input to the model. The model


consists of four components, all of which contribute to predicted tension. Of most interest here is the basic pitch space, a theorized version of empirical measures of tonal hierar-chies. The basic pitch space, shown in Fig. 3.2a, consists of five levels: the chromatic scale at the lowest level, the diatonic scale on the next level, then the triad level, the tonic-dominant level, and finally the tonic level.

To make quantitative predictions of the amount of tension caused by each event in the music, it is necessary to find a way of computing the distances between events. The distance between events is the sum of three numbers. The first, and most relevant here, is the number of changes that are needed to transform the basic pitch space for the first event into the basic pitch space for the second event. Take as an example the distance between I chord in C major (C major) and the vi chord (the chord built on the sixth tone of the F major scale, d minor). The basic pitch space for the C major chord is shown in Fig. 3.2a and the basic pitch space for the vi chord in F major (3 minor) is shown in Fig. 3.2b. There is one change on the diatonic level: B is changed to Bb. On the levels above this, C was represented on all five levels and is now represented on two levels (a difference of 3); E, which was represented on three levels is now represented on two levels (a difference of 1); and G, which was represented on four levels is now represented on two levels (a differ-ence of 2). This gives a total of 7 for the number of changes to transform one basic pitch space into the other. To these are added the distances between chords on the circle of fifths for chords (2, that is, C to G, G to d) and the distance between keys (1, that is, C major to F major), for a total tension value of 10.

a

b

(a) Octave (root) level: X

(b) Fifths level: X X

(c) Triad level: X X X

(d) Diatonic level: X X X X X X X

(e) Chromatic level: X X X X X X X X X X X X

C C# D D# E F F# G G# A Bb B

(a) Octave (root) level: X

(b) Fifths level: X X

(c) Triad level: X X X

(d) Diatonic level: X X X X X X X

(e) Chromatic level: X X X X X X X X X X X X

C C# D D# E F F# G G# A Bb B

Fig. 3.2 (a) The basic pitch space for the tonic triad chord (I) in C major. (b) The basic pitch space for the chord built on the second scale tone (ii) of F major. Seven elements change between the two basic spaces (See Lerdahl 2001; Lerdahl and Krumhansl 2007)


The distances as computed above are then applied to the prolongational structure, a tree representation that shows the way in which events in the music are related to one another. This is an instance of what Bharucha (1984) called an event hierarchy, as defined in the preceding text. The tree representation shows links that occur between events. It is important to note that these are sometimes nonadjacent in the music as shown in Fig. 3.3. The tree shows a link between the V chord (dominant, the major chord built on the fifth scale degree) and the I chord (tonic, the major chord built on the first scale degree, or tonic) although the ii chord (the minor chord built on the second scale degree) intervenes. In a tree such as this, the tension value for an event is the sum of the distances between the events at all levels up to the root of the tree. In Fig. 3.3, the ii chord is subordinate to the V chord and this is subordinate to the I chord at the root of the tree. The total distance for the ii chord from the root of the tree (its tension) is its distance from the V chord, plus the distance of the V chord from the I chord, with these distances computed as just described. To this value is added two other numbers, the surface dissonance (of each event) and attraction (between successive events) as specified in the model. This is the predicted tension.

To date, five studies have found empirical support for the model’s account of tension (Bigand et al. 1996; Krumhansl 1996; Cuddy and Smith 2000; Smith and Cuddy 2003; Lerdahl and Krumhansl 2007). In the empirical studies, participants were asked to judge tension (either continuously as the music is played or, for the shorter segments, stopped at each successive point) and these judgments were com-pared with the model’s quantitative predictions. Lerdahl and Krumhansl (2007) is the most extensive test of the model, using five excerpts written in quite different styles. The styles included traditional tonal-harmonic music, highly chromatic dia-tonic music, and two excerpts that might be analyzed in other scales, the six-tone hexatonic scale and the eight-tone octatonic scale. The predictions of the model

I ii V

Nonadjacent dependency

Fig. 3.3 Nonadjacent dependencies in Lerdahl’s (2001) prolongation structure. The V chord links to the I chord even though the ii chord intervenes. The tension of the ii chord is computed as the distance between the ii chord and the V chord plus the distance between the V chord and the I chord


were strongly confirmed for all the excerpts. In some cases, modifications of the prolongational (tree) structure achieved a better fit between model and data, but the modifications were principled in terms of the pitch space model.

What is most relevant here is that the distance calculations in the Lerdahl (2001) model required no modification. Recall that these distances were calculated using the basic space that has essentially the same structure as the empirically measured tonal hierarchies. It is not obvious a priori that this would be the case. The experimental studies on tone, chord, and key distances on which the model is based tend to use short, schematic materials (e.g., scales, chords, chord cadences). These materials lack relationships found in extended musical excerpts between nonadjacent events as described in the prolongational structure. In some cases the nonadjacent events are quite distant in the music. Nonetheless, the distances calculated along the branches of the trees successfully predicted the tension judgments. Thus, it is notable that the pitch space distance, calculated from the basic pitch space, extends to complex segments of musical compositions. In addition, the judgment of tension in these experiments is quite different from the more direct judgments of structure used in the basic experimental studies as described earlier.

This is an example of a deeply theorized proposal stimulated in part by the empirical work on tonal hierarchy. It makes predictions that can be tested against data, which may lead to refinements of the theory. Other music theoretic models based on the empirical results on tonal hierarchy and related work may follow.

3.5 Tonal Hierarchy Theory and Implications for Cognitive Science

In this chapter, a theory of tonal hierarchies has been presented. It is a theory in the sense that it summarizes a large body of empirical evidence and makes predictions for empirical study. The theory asserts three propositions, one concerning the psychological status of the tonal hierarchy, one concerning the musical status of the hierarchy, and the third concerning the relationship between these subjective and objective descriptions of the tonal hierarchy.

The first proposition is that tonal hierarchies have psychological reality. Psychological interest in tonal hierarchies grew out of music-theoretic descriptions of Western tonal-harmonic music, the style that dominates Western music. In this style, certain tones are identified as more prominent, stable, and structurally significant than others. Thus, according to the theory, music establishes a hierarchy of tones. In Western music, the hierarchy is headed by the tonic. Other tones follow: those in the tonic triad, then the other scale tones, and finally the nonscale tones. A wide range of experimental methods have been developed to test whether the proposed hierarchies can be elicited in psychological experiments. The psychological reality of the tonal hierarchy is supported by converging evidence from numerous experi-mental studies of cognition, development, learning, neuropsychology, and cross-cultural psychology. Among its manifestations are tonal hierarchy’s effects on


memory, the sense of stability and instability, the choice of phrase endings, the perceived relations between pitches, and the generation of melodic expectations.

Cross-cultural comparisons proved particularly important for sharpening the questions involved. For music from different cultures, listeners did not produce the Western tonal hierarchy but rather produced a tonal hierarchy consistent with the hierarchy of each musical culture. This general result argues against the view that tonal hierarchies can be explained by the harmonic structure (overtones) of complex tones. The argument acknowledges the fact that harmonic structure across cultures typically contains the perceptually privileged intervals of the octave and the perfect fifth. However, and significantly, if harmonic structure were the sole determinant of the tonal hierarchy then tonal hierarchies would not exhibit this kind of cross-cultural variability. Rather, these variations suggested that the tonal hierarchy is cognitive in origin and that it is internalized in large part through experience with music.

The second proposition is that tonal hierarchies are musical facts. As indicated earlier, the initial impulse for the empirical studies was the description in music theory of hierarchies of tones in tonal-harmonic music. However, this claim resides in a complex theory of tonal-harmonic music and technical vocabulary that would be unknown to nonmusicians. A musical property more readily perceptible is needed to explain the generality of the experimental results just summarized. The property that was identified was the relative emphasis given tones in the music, the assumption being that tones high in the tonal hierarchy would be emphasized on the surface of the music. An analytic method was needed for describing the relative salience of tones. It was quantified by determining the frequencies and/or durations of each tone in the music. A number of note-count studies have analyzed large corpora of Western tonal-harmonic music. The results vary somewhat, depending on the particular pieces in the corpus, but at a broad level the results correspond to the empirically measured tonal hierarchies. More narrowly focused studies con-sider particular styles of music or the way a composer writes music to make the tonal hierarchy evident. Analysis of styles other than tonal-harmony also shows surface emphasis by frequency and duration. This line of research has been aided by computer encoding of large numbers of pieces and tools to conduct such studies of tone distributions.

The evidence just described for the psychological reality of tonal hierarchies and their status as musical facts raises the fundamental question as to their relationship. This issue is articulated by the third proposition of the theory: statistically frequent patterns in music, and the consequent salience of certain tones, enable listeners to orient to tonal hierarchies. Moreover, the proposition predicts that listeners rapidly adapt to style-appropriate tonal hierarchies even if the style is unfamiliar.

That listeners use tone distributions to form a tonal hierarchy has now been shown by experiments with a wide range of styles. Probe tone judgments evoked in Western listeners by simple schematic key-defining contexts closely match the distribution of tones in tonal-harmonic music. Westerners exposed to music of other cultures, that of India and to some extent Bali, demonstrate that sensitivity to tone distributions in the contexts induces rapid assimilation of the relevant tonal hierarchies. Examination of


the music showed that tones high in the hierarchy tended to be repeated more often and for longer durations. In studies of nontonal Western music, sensitivity to tone distributions is also observed, but with some qualifications, especially in the case of 12-tone serial music. Finally, the correspondence between tone distributions and the perceived tonal hierarchy is also found when the distribution of tones is manipulated experimentally according to novel schemes. Some of the results suggest that musi-cians are somewhat more sensitive to the distributions of tones than nonmusicians. Finally, computational models provide additional supporting evidence. It has been shown that neural network models are able to abstract tonal hierarchies (as well as harmonic and key relations) from both schematic and more realistic musical inputs. In addition, processing of statistical properties of the musical surface (the frequencies of occurrence of tones and tone combinations) has provided the basis for successful modeling of tonality induction.

In sum, there is a strong basis in experimental results to support the first proposi-tion of the theory, that tonal hierarchies have psychological reality. A metric of musical tonal hierarchies, the distribution of tone frequencies and/or duration, matches well the psychological tonal hierarchy. Thus, tonal hierarchies are musical facts, which is the second proposition of the theory. Finally, building on these two first propositions, the third proposition is that sensitivity to the distribution of tones in the music enables listeners to abstract the tonal hierarchy. The evidence supports this, and also that listeners can adapt quite rapidly to unfamiliar styles.

Viewed from a broader cognitive science perspective, the literature reviewed here demonstrates the operation of two general psychological principles. The first is the existence of a frame of reference to guide perception and cognition that takes the form of a hierarchy of pitches. Other perceptual and cognitive domains contain reference points; these reference points provide an economical description of the domain in question. In the context of music, the tones high in the hierarchy serve as reference points with respect to which other tones and chords are efficiently encoded and accurately remembered.

Despite the common principle of reference points, tonal hierarchies appear to be unique to music. Nothing analogous appears, for example, in language or in other perceptual domains. This raises the possibility that tonal hierarchies are especially important in music because most listeners (those without absolute pitch, the ability to name tones in isolation) process music relatively. In other words, musical tones do not have inherent qualities that are invariant across contexts. Instead, pitches are heard in context, and related to one another in that context. The tonal hierarchy provides a stable framework for establishing these relationships. Moreover, the representation of the hierarchy may become isolable at the neural level so that it may be selectively lost or spared in brain pathology.

The second general psychological principle that underlies tonal hierarchies is a mechanism for extracting environmental regularities through sensitivity to distribu-tions of tones and tone combinations. This principle came into focus as the empirical question shifted from establishing the psychological reality of tonal hierarchies to investigating whether tonal hierarchies are learned and, if so, how. The evidence pointed to the involvement of sensitivity to tone distributions, specifically how the


tones are emphasized by frequency and duration. These measures could be readily quantified independently of the style. They were found to match well the empirical measures of tonal hierarchy. This surface emphasis appeared to be accessible to nonmusicians as seen by their judgments closely following the tone distributions. The results for listeners familiar with the style also reflected the tone distributions in addition to the influences of style-specific knowledge. This suggests that tone distri-butions and style-specific knowledge are complementary, rather than conflicting.

Music appears to be very distinct from other perceptual and cognitive domains. Moreover, a vast variety of different styles of music can be found across cultures and historical periods. This suggests it would be difficult to draw on other domains of psychological study or to make generalizations across musical styles. However, when viewed in terms of underlying cognitive principles, the commonalities with other domains are revealed at a deep level and the diversity of musical styles can be better understood.

References

Aarden B (2003) Dynamic melodic expectancy. Unpublished doctoral dissertation, School of Music, Ohio State University.

Auhagen W, Vos PG (2000) Experimental methods in tonality induction research: a review. Music Percept 17:417–436.

Ayotte J, Peretz I, Hyde K (2002) Congenital amusia: a group study of adults afflicted with a music-specific disorder. Brain 125:238–251.

Besson M, Faïta F (1995) An event-related potential (ERP) study of musical expectancy: compari-son of musicians with nonmusicians. J Exp Psychol Hum Percept Perform 21:1278–1296.

Besson M, Faïta F, Peretz I, Bonnel AM, Requin J (1998) Singing in the brain: independence of lyrics and tunes. Psychol Sci 9:494–498.

Bharucha JJ (1984) Event hierarchies, tonal hierarchies and assimilation: a reply to Deutsch and Dowling. J Exp Psychol Gen 113:421–425.

Bharucha JJ, Krumhansl CL (1983) The representation of harmonic structure in music: hierarchies of stability as a function of context. Cognition 13:63–102.

Bigand E, Poulin-Charronnat B (2006) Are we “experienced listeners”? A review of the musical capacities that do not depend on formal musical training. Cognition 100:100–130.

Bigand E, Parncutt R, Lerdahl F (1996) Perception of musical tension in short chord sequences: the influence of harmonic function, sensory dissonance, horizontal motion and musical train-ing. Percept Psychophys 58:125–141.

Boltz M (1989a) Perceiving the end: effects of tonal relationships on melodic completions. J Exp Psychol Hum Percept Perform 15:749–761.

Boltz M (1989b) Rhythm and “good endings”: effects of temporal structure on tonality judgments. Percept Psychophys 46: 9–17.

Brown H, Butler B, Jones MR (1994) Musical and temporal influences on key discovery. Music Percept 11:371–407.

Castellano MA, Bharucha JJ, Krumhansl CL (1984) Tonal hierarchies in the music of North India. J Exp Psychol Gen 113:394–412.

Cohen AJ (1991) Tonality and perception: musical scales primed by excerpts from the Well-tempered Clavier of JS Bach. Psychol Res/Psychol Forsch 53:305–314.

Cohen AJ (2000) Development of tonality induction: plasticity, exposure, and training. Music Percept 17:437–459.


Cuddy LL (1997) Tonal relations. In: Deliège I, Sloboda J (eds), Perception and Cognition of Music. Hove, Sussex: Taylor & Francis, pp. 329–352.

Cuddy LL (2000) Perception and representation of musical structure. Paper presented at the XXVII International Congress of Psychology, Stockholm, July 23–28.

Cuddy LL, Badertscher, B (1987) Recovery of the tonal hierarchy: some comparisons across age and levels of musical experience. Percept Psychophys 41:609–620.

Cuddy LL, Duffin JM (2005) Music, memory, and Alzheimer’s disease: is music recognition spared in dementia and how can it be assessed? Med Hypotheses 64:229–235.

Cuddy LL, Lunney CA (1995) Expectancies generated by melodic intervals: perceptual judgments of melodic continuity. Percept Psychophys 57:451–462.

Cuddy LL, Smith NA (2000) Perception of tonal pitch space and tonal tension. In: Greer D (ed), Musicology and Sister Disciplines: Past Present Future. Oxford: Oxford University Press, pp. 47–59.

Cuddy LL, Cohen AJ, Mewhort DJK (1981) Perception of structure in short melodic sequences. J Exp Psychol Hum Percept Perform 7:869–883.

Cuddy LL, Balkwill L-L, Peretz I, Holden RR (2005) Musical difficulties are rare: a study of “tone deafness” among university students. Ann NY Acad Sci 1060:311–324.

Dalla Bella S, Peretz I (1999) Music agnosias: selective impairments of music recognition after brain damage. J New Music Res 28:209–216.

DeVoto M (1986) Tonality. In: Randel DM (ed), The New Harvard Dictionary of Music. Cambridge, MA: Belknap, pp. 862–863.

Dowling WJ (1978) Scale and contour: two components of a theory of memory for melodies. Psychol Rev 85:341–354.

Dowling WJ (1999) The development of music perception and cognition. In: Deutsch D (ed), The Psychology of Music (2nd ed.) San Diego: Academic Press, pp. 603–626.

Drayna D, Manichaikul A, de Lange M, Snieder H, Spector T (2001) Genetic correlates of musical pitch recognition in humans. Science 291:1969–1972.

Fornazzari L, Castle T, Nadkarni S, Ambrose M, Miranda D, Apanasiewicz N, et al. (2006) Preservation of episodic musical memory in a pianist with Alzheimer disease. Neurology 66: 610–611.

Frankland BW, Cohen AJ (1990) Expectancy profiles generated by major scales: group differ-ences in ratings and reaction time. Psychomusicology 9:173–192.

Garner WR, Hake HW, Eriksen CW (1956) Operationism and the concept of perception. Psychol Rev 63:149–159.

Handel S (1989) Listening: An Introduction to the Perception of Auditory Events. Cambridge, MA: MIT Press.

Hughes M (1977) A quantitative analysis. In: Yeston M (ed), Readings in Schenker Analysis and Other Approaches. New Haven: Yale University Press, pp. 144–164.

Huron D (1993) Chordal-tone doubling and the enhancement of key perception. Psychomusicology 12:154–171.

Huron D (2006) Sweet Anticipation: Music and the Psychology of Expectation. Cambridge, MA: MIT Press.

Huron D, Parncutt R (1993) An improved model of tonality perception incorporating pitch salience and echoic memory. Psychomusicology 12:154–171.

Hyde K, Peretz I (2004) Brains that are out of tune but in time. Psychol Sci 15:356–360.Janata P, Reisberg D (1988) Response-time measures as a means of exploring tonal hierarchies.

Music Percept 6:161–172.Järvinen T (1995) Tonal hierarchies in jazz improvisation. Music Percept 12:415–437.Jordan DS, Shepard RN (1987) Tonal schemas: evidence obtained by probing distorted musical

scales. Percept Psychophys 41:489–504.Kessler EJ, Hansen C, Shepard RN (1984) Tonal schemata in the perception of music in Bali and

in the West. Music Percept 2:131–165.Knopoff L, Hutchinson W (1983) Entropy as a measure of style: the influence of sample length.

J Music Theory 27:75–97.


Krumhansl CL (1979) The psychological representation of musical pitch in a tonal context. Cogn Psychol 11:346–374.

Krumhansl CL (1985) Perceiving tonal structure in music. Am Sci 73: 371–378.Krumhansl CL (1987) Tonal and harmonic hierarchies. In: Sundberg J (ed), Harmony and

Tonality. Stockholm: Royal Swedish Academy, pp. 13–32.Krumhansl CL (1990a) Cognitive Foundations of Musical Pitch. New York: Oxford University Press.Krumhansl CL (1990b) Tonal hierarchies and rare intervals in music cognition. Music Percept

7:309–324.Krumhansl CL (1991) Melodic structure: theoretical and empirical descriptions. In: Sundberg J

(ed), Music, Language, Speech and Brain. London: Macmillan.Krumhansl CL (1995a) Effects of musical context on similarity and expectancy. Systematische

Musikwissenschaft (Systematic Musicology) 3:211–250.Krumhansl CL (1995b) Music psychology and music theory: Problems and prospects. Music

Theory Spectrum 17:53–80.Krumhansl CL (1996) A perceptual analysis of Mozart’s piano sonata K. 282: segmentation ten-

sion and musical ideas. Music Percept 13:401–432.Krumhansl CL (2002) Music: a link between cognition and emotion. Curr Dir Psychol Sci 11:45–50.Krumhansl CL, Agres KA (2008) Musical expectancy: the influence of musical structure on emo-

tional response. Brain Behav Sci 31:584–585.Krumhansl CL, Keil FC (1982) Acquisition of the hierarchy of tonal functions in music. Mem

Cogn 10:243–251.Krumhansl CL, Kessler EJ (1982) Tracing the dynamic changes in perceived tonal organization in

a spatial representation of musical keys. Psychol Rev 89:334–368.Krumhansl CL, Schmuckler MA (1986) The Petroushka chord: a perceptual investigation. Music

Percept 4:153–184.Krumhansl CL, Shepard RN (1979) Quantification of the hierarchy of tonal functions within a

diatonic context. J Exp Psychol Hum Percept Perform 5:579–594.Krumhansl CL, Toiviainen P (2001) Tonal cognition. Ann NY Acad Sci 930:77–91.Krumhansl CL, Sandell GJ, Sergeant DC (1987) The perception of tone hierarchies and mirror

forms in twelve-tone serial music. Music Percept 5:31–78.Krumhansl CL, Louhivuori J, Toiviainen P, Järvinen T, Eerola T (1999) Melodic expectation in

Finnish spiritual folk hymns: convergence of statistical behavioral and computational approaches. Music Percept 17:151–195.

Krumhansl CL, Toivanen P, Eerola T, Toiviainen P, Järvinen T, Louhivuori J (2000) Cross-cultural music cognition: cognitive methodology applied to North Sami yoiks. Cognition 76:13–58.

Lamont A, Cross I (1994) Children’s cognitive representations of musical pitch. Music Percept 12:27–55.

Lantz ME, Cuddy LL (1998) Total and relative duration as cues to surface structure in music. Can Acoust 26:56–57.

Lantz ME, Kilgour A, Nicholson KG, Cuddy LL (2003) Judgments of musical emotion following right hemisphere damage. Brain Cogn 51:190–191.

Leman M (2000) An auditory model of the role of short-term memory in probe-tone ratings. Music Percept 17:481–509.

Lerdahl F (1988) Tonal pitch space. Music Percept 5:315–349.Lerdahl F (1996) Calculating tonal tension. Music Percept 13:319–363.Lerdahl F (2001) Tonal Pitch Space. New York: Oxford University Press.Lerdahl F (2009) Genesis and architecture of the GTTM project. Music Percept 26:187–194.Lerdahl F, Krumhansl CL (2007) Modeling tonal tension. Music Percept 24:329–366.Liégeois-Chauvel C, Peretz I, Babaï M, Laguitton V, Chauvel P (1998) Contribution to different

cortical areas in the temporal lobes to music processing. Brain 121:1853–1867.Longuet-Higgins HC, Steedman MJ (1971) On interpreting Bach. Mach Intell 6:221–241.Marin OSM, Perry DW (1999) Neurological aspects of music perception and performance. In:

Deutsch D (ed), The Psychology of Music (2nd ed.) San Diego: Academic Press, pp. 653–724.Marmel F, Tillmann B (2009) Tonal priming beyond tonics. Music Percept 26:211–221.


Meyer LB (1956) Emotion and Meaning in Music. Chicago: University of Chicago Press.Miyazaki K (1989) Absolute pitch identification: effects of timbre and pitch region. Music Percept

7:1–14.Narmour E (1990) The Analysis and Cognition of Basic Melodic Structures: The Implication-

Realization Model. Chicago: University of Chicago Press.Oram N, Cuddy LL (1995) Responsiveness of Western adults to pitch-distributional information

in melodic sequences. Psychol Res 57:103–118.Palmer C, Krumhansl CL (1987a) Independent temporal and pitch structures in perception of

musical phrases. J Exp Psychol Hum Percept Perform 13:116–126.Palmer C, Krumhansl CL (1987b) Pitch and temporal contributions to musical phrase perception:

effects of harmony, performance timing and familiarity. Percept Psychophys 41:505–518.Parncutt R (1988) Revision of Terhardt’s psychoacoustical model of the root(s) of a musical chord.

Music Percept 6:65–94.Parncutt R (1989) Harmony: A Psychoacoustical Approach. Berlin: Springer-Verlag.Parncutt R (1994) Template-matching models of musical pitch and rhythm perception. J New

Music Res 23:145–167.Patel AD (2008) Music, Language, and the Brain. Oxford: Oxford University Press.Peretz I (1993a) Auditory agnosia: a functional analysis. In: McAdams S, Bigand E (eds),

Thinking in Sound: The Cognitive Psychology of Human Audition. Oxford: Oxford University Press, pp 199–230.

Peretz I (1993b) Auditory atonalia for melodies. Cogn Neuropsychol 10:21–56.Peretz I (1996) Can we lose memory for music? A case of music agnosia in a nonmusician. J Cogn

Neurosci 8:481–496.Peretz I (2008). Musical disorders: from behavior to genes. Curr Dir Psychol Sci 17:329–333.Peretz I, Kolinsky R, Tramo M, Labrecque R, Hublet C, Demeurisse G, Belleville S (1994)

Functional dissociations following bilateral lesions of auditory cortex. Brain 117:1283–1301.Peretz I, Belleville S, Fontaine S (1997) Dissociations entre musique et langage après atteinte

cérébrale: un nouveau cas d’amusie sans aphasie. Can J Exp Psychol 51:354–368.Peretz I, Gosselin N, Tillmann B, Cuddy LL, Gagnon B, Trimmer CG, Paquette S, Bouchard B

(2008) On-line identification of congenital amusia. Music Percept 25:331–343.Perlman M, Krumhansl CL (1996) An experimental study of internalized interval standards of

Javanese and Western musicians. Music Percept 14:95–116.Piston W (1987) Harmony (revised and expanded by M DeVoto). New York: WW Norton.Rosch E (1975) Cognitive reference points. Cogn Psychol 7:532–47.Rosch E (1978) Principles of categorization. In: Rosch E, Lloyd BB (eds), Cognition and

Categorization. Hillsdale, NJ: Lawrence Erlbaum.Rosch E (1979) On the internal structure of perceptual and semantic categories. In: Moore TE (ed),

Cognitive Development and the Acquisition of Language. New York: Academic Press.Rosch E, Mervis CB (1975) Family resemblances: studies in the internal structure of categories.

Cogn Psychol 7:573–605.Rowe R (2000) Key induction in the context of interactive performance. Music Percept 17:511–530.Saffran JR, Griepentrog GJ (2001) Absolute pitch in infant auditory learning: evidence for devel-

opmental reorganization. Dev Psychol 37:74–85.Saffran JR, Aslin RN, Newport EL (1996a) Statistical learning by 8-month-old infants. Science

274:1926–1928.Saffran JR, Newport EL, Aslin RN (1996b) Word segmentation: the role of distributional cues.

J Mem Lang 35:606–621.Saffran JR, Newport EL, Aslin RN, Tunick RA, Barrueco S (1997) Incidental language learning:

listening (and learning) out of the corner of your ear. Psychol Res 8:101–105.Saffran JR, Johnson EK, Aslin RN, Newport EL (1999) Statistical learning of tone sequences by

human infants and adults. Cognition 70:27–52.Schmuckler MA (1989) Expectation in music: investigation of melodic and harmonic processes.

Music Percept 14:295–318.Shepard RN (1964) Circularity in judgments of relative pitch. J Acoust Soc Am 36:2346–2353.


Shuter-Dyson R (1999) Musical ability. In: Deutsch D (ed), The Psychology of Music (2nd ed) San Diego: Academic Press, pp. 627–652.

Smith NA, Cuddy LL (2003) Perceptions of musical dimensions in Beethoven’s Waldstein Sonata: an application of tonal pitch space theory. Musicae Scientiae 7:7–34.

Smith NA, Schmuckler MA (2004) The perception of tonal structure through the differentiation and organization of pitches. J Exp Psychol Hum Percept Perform 30:268–286.

Speer JR, Meeks PU (1985) School children’s perception of pitch in music. Psychomusicology 5:49–56.

Steinke WR, Cuddy LL, Peretz I (1994) Dissociation of music and cognitive abstraction abilities in normal and neurologically impaired subjects. In: Proceedings of the 3rd International Conference on Music Perception and Cognition, Liège, Belgium pp. 425–426.

Steinke WR, Cuddy LL, Holden RR (1997) Dissociation of musical tonality and pitch memory from nonmusical cognitive abilities. Can J Exp Psychol 51:316–334.

Steinke WR, Cuddy LL, Jakobson LS (2001) Dissociations among functional subsystems govern-ing melody recognition after right hemisphere damage. Cogn Neuropsychol 18:411–437.

Temperley D (1999) What’s key for key? The Krumhansl-Schmuckler key-finding algorithm reconsidered. Music Percept 17:65–100.

Temperley D (2001) The Cognition of Basic Musical Structures. New York: Oxford University Press.

Temperley D (2007) Music and Probability. Cambridge, MA: M.IT Press.Temperley D, Marvin EW (2008) Pitch-class distribution and the identification of key. Music

Percept 25:193–212.Terhardt E (1979) Calculating virtual pitch. Hear Res 1:155–182.Terhardt E, Stoll G, Seewann M (1982a) Pitch of complex signals according to virtual-pitch

theory. Test examples and predictions. J Acoust Soc Am 71:671–678.Terhardt E, Stoll G, Seewann M (1982b) Algorithm for extraction of pitch and pitch salience from

complex tonal signals. J Acoust Soc Am 71:679–388.Thompson WF (2008) Music, Thought, and Feeling: Understanding the Psychology of Music.

New York: Oxford University Press.Thompson WF, Cuddy LL (1997) Music performance and the perception of key. J Exp Psychol

Hum Percept Perform 23:116–135.Thompson WF, Mor S (1992) A perceptual investigation of polytonality. Psychol Res 54:60–71.Thompson WF, Cuddy LL, Plaus C (1997) Expectancies generated by melodic intervals: evalua-

tion of principles of melodic implication in a melody completion task. Percept Psychophys 59:1069–1076.

Tillmann B, Bharucha JJ, Bigand E (2000) Implicit learning of tonality: a self-organizing approach. Psychol Rev 107:885–913.

Toiviainen P, Krumhansl CL (2003) Measuring and modeling real-time responses to music: tonal-ity induction. Perception 32:741–766.

Trainor LJ, Trehub SE (1994) Key membership and implied harmony in Western tonal music: developmental perspectives. Percept Psychophys 56:125–132.

Trehub SE (2000) Human processing predispositions and musical universals. In: Wallin NL, Merker B, Brown S (eds), The Origins of Music. Cambridge, MA: MIT Press.

Trehub SE, Trainor LJ (1993) Listening strategies in infancy: the roots of music and language development. In: McAdams S, Bigand E (eds), Thinking in Sound: The Cognitive Psychology of Human Audition. Oxford: Oxford University Press, pp 278–327.

Trehub SE, Schellenberg EG, Hill DS (1997) The origins of music perception and cognition: a developmental perspective. In: Deliège I, Sloboda J (eds), Perception and Cognition of Music. Hove, Sussex: Taylor & Francis, pp. 103–128.

Vanstone AD, Cuddy LL (2010) Musical memory in Alzheimer disease. Aging, Neuropsychol Cogn 17:108–128.

Vos PG, Leman M. (2000) Guest editorial: tonality induction. Music Percept 17:401–544.Whitfield IC, Evans EF (1965) Responses of auditory cortical neurons to stimuli of changing

frequency. J Physiol 28:655–672.


Yonelinas AP, Hockley WE, Murdock BB (1992) Tests of the list-strength effect in recognition memory. J Exp Psychol Learn Mem Cogn 18:345–355.

Youngblood JE (1958) Style as information. J Music Theory 2:24–35.Zenatti A (1993) Children’s musical cognition and taste. In Tighe TJ, Dowling WJ (eds),

Psychology of Music: The Understanding of Melody and Rhythm. Hillsdale, NJ: Lawrence Erlbaum, pp. 177–196.

a theory of tonal hierarchies in music - the music cognition

Documents